Move from Perturbed scheme to exponential weighting average.

Published in: CoRR (2015)

Keyphrases

neural network
similarity measure
hash functions
learning scheme
efficient computation
representation scheme
optimization scheme